Forums - inception v3 does not predict correct using qtimlesnpe

3 posts / 0 new
Last post
inception v3 does not predict correct using qtimlesnpe
prabukumar
Join Date: 8 Feb 21
Posts: 19
Posted: Mon, 2021-10-18 02:22

Hi,

I am using the inception_v3.dlc which is generated using this script($SNPE_ROOT/models/inception_v3/scripts/setup_inceptionv3.py) . This gives correct prediction in snpe-net-run binary with the images generated when setup_inception_v3.py which is mentioned below.

snpe-1.50.0.2622/models/inception_v3/data/cropped$ ls
chairs.jpg  chairs.raw  notice_sign.jpg  notice_sign.raw  plastic_cup.jpg  plastic_cup.raw  raw_list.txt  trash_bin.jpg  trash_bin.raw

But when I show the same images using the camera and qtimlesnpe plugin , I get wrong predictions.(for cabbage I get hammer .etc) The pipeline I use is mentioned below.

gst-launch-1.0 -e qtiqmmfsrc ! video/x-raw, format=NV12, width=1280, height=720, framerate=30/1, camera=0 ! qtimlesnpe model=/data/local/tmp/inception_v3/inception_v3.dlc labels=/data/imagenet_slim_labels.txt postprocessing=classification ! qtioverlay !  qtivtransform rotate=1 ! waylandsink async=true sync=false fullscreen=true enable-last-sample=false

 

Also I need to test my custom images with snpe-net-run binary .What is the resolution and format of the image to be fed to the model using snpe-net-run for inception_v3.dlc ?

 



 

  • Up0
  • Down0
ap.arunraj
Join Date: 20 Apr 20
Posts: 21
Posted: Tue, 2021-10-19 06:40

Hello prabukumar,
You can refer $SNPE_ROOT/
models/inception_v3/scripts/create_inceptionv3_raws.py for the preprocessing required for the inception v3 model used in the tutorial.

  • Up0
  • Down0
prabukumar
Join Date: 8 Feb 21
Posts: 19
Posted: Mon, 2021-10-25 01:35

Hi arunraj,

Thanks for your reply.

Could you please provide some pointers to a working model which works with qtimelsnpe. Is there any pretrained dlc file which works with the qtimlesnpe . Also could you please provide the pipeline ?

 

  • Up0
  • Down0
or Register

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.